Epidemiological studies suggest that certain vitamin supplements may r
educe the risk of some cancers. However, observational studies can be
compromised by confounding, because supplement use is related to other
factors that affect cancer risk. The purpose of this paper is to iden
tify cancer-related behaviors that could confound studies of the assoc
iations between vitamin supplement use and cancer risk. Data are from
a random digit dial survey to monitor cancer risk behavior in adults i
n Washington State (n = 1449). Unconditional logistic regression was u
sed to examine whether regular supplement users were more likely to pr
actice other cancer-related behaviors than nonusers, after adjustment
for age, education, and smoking. Among women, supplement users were mo
re likely to have had a sigmoidoscopy [odds ratio (OR), 2.3; 95% confi
dence interval (CI), 1.2-4.5], hemoccult (OR, 2.3; CI, 1.5-3.5), or ma
mmogram (OR, 1.5; CI, 1.0-2.1) in the past 2 years. Among men, supplem
ent users were twice as likely to have had a prostate-specific antigen
test (OR, 2.2; CI, 1.3-3.7) and to regularly take aspirin (OR, 1.7; C
I, 1.1-2.6). Supplement users were statistically significantly more li
kely to exercise regularly, eat four or more servings of fruits and ve
getables per day, follow a low-fat diet pattern, and believe in a conn
ection between diet and cancer. The association was especially strong
for fruits and vegetables (women, OR, 1.9; and CI, 1.3-2.6; men, OR, 2
.4; CI, 1.6-3.8). Those investigating the benefits and risks of vitami
n and mineral supplements need to be aware of the lifestyle characteri
stics of supplement users to assess the potential for bias in their st
udies.